CognitiveCanvas Turns AI Chat Memory into Auditable Graphs, Cutting Tokens 92.2%
JeremyCMorgan · x · 2026-10-30
- CognitiveCanvas is an interactive tool (with an accompanying arXiv paper, no public repo yet) that turns multi-session AI coding conversations from linear chat logs into inspectable, editable Goal / Fact / Code objects. The pitch isn't persistent memory — it's letting developers audit and correct the relationships shaping an agent's answers.
- Key components:
- Hierarchical Semantic Fact-Graph (HSFG): distills dialogues into structured artifacts with 92.2% token reduction while enabling multi-hop GraphRAG traversal;
- Continuous Temporal Drift Auditor: 60+ regex heuristics plus live calendar-anchored LLM verification to catch silent context poisoning from outdated APIs/packages, with 1-click self-correction;
- Spatial-Symbolic Prompt Compiler (SSPC): translates 2D direct-manipulation gestures (Voronoi clustering) into structured prompts, addressing prompt bloat and attention degradation.
- The paper targets three pain points: flat-vector retrieval failing multi-hop dependencies, temporal drift poisoning reused context, and attention degradation from raw conversation concatenation.
More from coding & agent
- 10-person Spawn ships 77,000 changes a month, spends 30x payroll on AI tokens — TAbrodi · 2026-10-30
- qwen-code v0.27.0 ships cross-platform CUA driver binaries and MCP payload filter — github-actions[bot] · 2026-10-30
- Google Knowledge Catalog goes GA with semantic model ingestion from six sources — AiJohnAllen · 2026-10-30
- Lean-checked: an AI agent improved a formalized RG bound from 10^-10^10017 to 0.00034 — _onionesque · 2026-10-30
- Codex issue reports Dot subagent losing authorizations and stalling delegated work — ingChristopherCV · 2026-10-30
- How Agno's 1,428-line Agent class stays thin: a façade pattern teardown — Mahmoud_Zalt · 2026-10-30